CRITERIA FOR DELETING AN ITEM

[Ken Smith-14]

Can any one assist?

 

When employing  reliability analysis and looking at the column for “alpha if item is deleted” I recall reading that one criteria is the alpha need to increase at least 0.05.  Can anyone inform me of the reference?

 

Also, is there a criteria for deleting an item?

 

Any assistance is appreciated.

 

Thank you,

 

Ken--------------

 

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Ken Smith, PhD, MAPS, MAPA

Associate Professor

Australian Catholic University

Faculty of Education

School of Education

Melbourne Campus (St Patrick's)

Locked Bag 4115

Fitzroy MDC, Victoria, Australia 3065

(61 + 3) 9953-3257 (Tel)

(61 + 3) 9953-3495 (Fax)

[hidden email]

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

ABN 15 050 192 660, CRICOS Reg. 00004G, 00112C, 00873F, 00885B

This correspondence is for the named person's use only.  It may contain confidential or

 legally privileged information or both.  No confidentiality or privilege is waived or lost by this transmission. 

If you receive this correspondence in error, please immediately delete it from your system and notify the sender. 

You must not disclose, copy or rely on any part of this correspondence if you are not the Intended recipient.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

 

[Mike]

Although there may be such a "mechanical rule" for reliability analysis,

I would caution one about using it.  Instead, I would ask one to consider

the following points:

 

(1)  In terms of construct validity, whatever construct(s) the items are

serving as empirical indicators of, how important to the construct are

items that have "lower" or "less than desirable" alpha?  If one treats

all of the items as having the same importance, then a mechanical rule

can be applied because, in essence, one is saying "I don't care what

aspect of the construct this item is measuring, it doesn't measure it

well enough".

 

(2)  Some items may represent certain ideas that underlie a construct

and a researcher may be concerned with whether these items have

desirable statistical properties.  If such an item reduces a scale's alpha,

then one might ask (a) is this item not as important as I thought it was

and/or (b) is this item worded in the best possible way, free of ambiguity

or multiple interpretations or can this item be re-worded/rewritten to make

it a better expression of the aspect of the construct that the researcher

is actually interested in?

 

Presumably the question below was asked because one engaged in evaluating

an instrument with a number of items measuring different aspects of a single

construct.  Was the selection of items primarily driven by theoretical concerns

or by statistical concerns (this is not a dichotomous issue but it does ask the

question of how well one knows and understands the phenomenon being measured).

 

An item on a scale should be dropped or kept because of the information it

provides the researcher/user of the instrument.  Even items wiht poor statistical

properties might be useful if they provide info that is thought to be important.

However, the concern would then be, if this aspect is so important, how can I

make the items measuring it into better measures?

 

-Mike Palij

New York University

[hidden email]

 

 

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Sunday, February 07, 2010 8:55 PM

Subject: CRITERIA FOR DELETING AN ITEM

Can any one assist?

 

When employing  reliability analysis and looking at the column for “alpha if item is deleted” I recall reading that one criteria is the alpha need to increase at least 0.05.  Can anyone inform me of the reference?

 

Also, is there a criteria for deleting an item?

 

Any assistance is appreciated.

 

Thank you,

 

Ken--------------

 

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Ken Smith, PhD, MAPS, MAPA

Associate Professor

Australian Catholic University

Faculty of Education

School of Education

Melbourne Campus (St Patrick's)

Locked Bag 4115

Fitzroy MDC, Victoria, Australia 3065

(61 + 3) 9953-3257 (Tel)

(61 + 3) 9953-3495 (Fax)

[hidden email]

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

ABN 15 050 192 660, CRICOS Reg. 00004G, 00112C, 00873F, 00885B

This correspondence is for the named person's use only.  It may contain confidential or

 legally privileged information or both.  No confidentiality or privilege is waived or lost by this transmission. 

If you receive this correspondence in error, please immediately delete it from your system and notify the sender. 

You must not disclose, copy or rely on any part of this correspondence if you are not the Intended recipient.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^